def performsSlidingWindowForecast(filename, minpercentile=5, step=30, input_window=3000, predic_window=30):
'''
Input window = 250 hours = 250*12 = 3000
look ahead window 60 samples = 5 hours = 720min/5 = 60
'''
data = np.genfromtxt(filename)
data = data/np.max(data)
minimum = np.percentile(data,minpercentile)
N = len(data)
result = []
print filename, "started..."
y = data[0:input_window].tolist()
model = HW_model.HW_model(y, minimum, 'additive')
for strIndex in range(0,N-input_window - predic_window, step):
if strIndex == 0:
model.fit()
else:
y = data[strIndex:strIndex+input_window].tolist()
model.update(y)
y_pred = model.predict(fc=predic_window)
result.append(y_pred)
f = filename.split('/')[-1]
fileutils.writeCSV("D:/Wikipage data/pageviews_hw/"+f, np.atleast_2d(result))
print filename, "complete!"
def performsSlidingWindowForecast(filename_eta_lmda, minpercentile=5, step=30, input_window=3000, predic_window=30):
'''
Input window = 250 hours = 250*12 = 3000
look ahead window 60 samples = 5 hours = 720min/5 = 60
'''
filename, curEta, curLmda = filename_eta_lmda
data = np.genfromtxt(filename, delimiter=',', skip_header=1, usecols=(1))
minimum = np.percentile(data,minpercentile)
N = len(data)
print filename, "started..."
curMachine = filename.split('/')[-1]
model = Fnn_model.Fnn_model(data=data, machineID = curMachine,netPath='../data/cpu2_networks/'+curMachine.replace(".csv",".xml"), eta=curEta, lmda=curLmda)
model.fit()
pred = []
# lastFc = 0
for p in range(input_window, len(data)-predic_window,predic_window):
fc = np.array(model.predict(predic_window)).flatten()
# fc[-1] = lastFc
# lastFc = fc[0]
fc[fc<0] = minimum
pred.append(fc)
model.update()
pred = np.array(pred)
f = filename.split('/')[-1]
fileutils.writeCSV("d:/data/cpu2_fnn/"+f, np.atleast_2d(pred))
print filename, "complete!"
def performsSlidingWindowForecast(filename,
minpercentile=5,
step=30,
input_window=3000,
predic_window=30):
'''
Input window = 250 hours = 250*12 = 3000
look ahead window 60 samples = 5 hours = 720min/5 = 60
'''
data = np.nan_to_num(np.genfromtxt(filename, delimiter=',', skip_header=1))
minimum = np.percentile(data[:, 1], minpercentile)
N = len(data[:, 1])
result = []
print filename, "started..."
for strIndex in range(0, N - input_window - predic_window, step):
if strIndex == 0:
y = data[:input_window, 1]
model = Norm_model.Norm_model(y)
model.fit()
else:
y = data[strIndex:strIndex + input_window, 1]
model.update(y)
y_pred = model.predict(predic_window)
y_pred[y_pred < 0] = minimum
result.append(y_pred)
f = filename.split('/')[-1]
fileutils.writeCSV("d:/data/cpu_norm_forecasts/" + f,
np.atleast_2d(result))
print filename, "complete!"
def extractMachineData():
datafiles = fileutils.getFilelist("D:/googleClusterData/clusterdata-2011-1/task_usage")
machineUsage = {}
startAt = 0;
for machine in machines:
machineUsage[machine] = []
for datafile in datafiles[startAt:]:
print datafile
for row in fileutils.getCsvRows(datafile):
curMachine = row[4]
if curMachine in machines:
machineUsage[curMachine].append(row)
for machine in machineUsage.keys():
if startAt == 0:
fileutils.writeCSV("d:/data/perMachine5/"+machine+".csv", machineUsage[machine])
else:
fileutils.writeCSV("d:/data/perMachine5/"+machine+".csv", machineUsage[machine], mode='ab')
startAt += 1
machineUsage.clear()
for machine in machines:
machineUsage[machine] = []
def performsSlidingWindowForecast(filename, minpercentile=5, step=30, input_window=3000, predic_window=30):
'''
Input window = 250 hours = 250*12 = 3000
look ahead window 60 samples = 5 hours = 720min/5 = 60
'''
#Wikidata
# data = np.genfromtxt(filename)
# data = data/np.max(data)
data = np.nan_to_num(np.genfromtxt(filename, delimiter=',', skip_header=1)[:,1]).ravel()
minimum = np.percentile(data,minpercentile)
N = len(data)
result = []
print filename, "started..."
for strIndex in range(0,N-input_window - predic_window, step):
if strIndex == 0:
y = data[:input_window]
model = AR_model.AR_model(y, order=30)
model.fit()
else:
y = data[strIndex:strIndex+input_window]
model.update(y)
y_pred = model.predict(predic_window)
y_pred[y_pred[:,0]<0,0] = minimum
result.append(y_pred[:,0])
f = filename.split('/')[-1]
fileutils.writeCSV("d:/data/"+TYPE+"_ar/"+f, np.atleast_2d(result))
print filename, "complete!"
def performsSlidingWindowForecast(filename, minpercentile=5, step=30, input_window=3000, predic_window=30):
'''
Input window = 250 hours = 250*12 = 3000
look ahead window 60 samples = 5 hours = 720min/5 = 60
'''
# Wikidata
# data = np.genfromtxt(filename)
# data = data/np.max(data)
data = np.nan_to_num(np.genfromtxt(filename, delimiter=',', skip_header=1)[:,1]).ravel()
minimum = np.percentile(data,minpercentile)
N = len(data)
result = []
max = np.max(data)
print filename, "started..."
for strIndex in range(0,N-input_window - predic_window, step):
if strIndex == 0:
y = data[:input_window]
model = Press_model.Press_model(y, maximum=max)
model.fit()
else:
# y = data[strIndex:strIndex+input_window,1]
y = data[input_window + strIndex - step:input_window + strIndex]
model.update(y)
y_pred = model.predict(predic_window)
y_pred[y_pred<0] = minimum
result.append(y_pred)
res = np.zeros((len(result),predic_window))
for i in range(len(result)):
res[i,:len(result[i])] = result[i]
f = filename.split('/')[-1]
fileutils.writeCSV("d:/data/"+TYPE+"_press/"+f, np.atleast_2d(res))
print filename, "complete!"
def performsSlidingWindowForecast(filename_eta_lmda, minpercentile=5, step=30, input_window=3000, predic_window=30):
'''
Input window = 250 hours = 250*12 = 3000
look ahead window 60 samples = 5 hours = 720min/5 = 60
'''
filename, curEta, curLmda = filename_eta_lmda
data = np.genfromtxt(filename, delimiter=',', skip_header=1, usecols=(1))
minimum = np.percentile(data,minpercentile)
N = len(data)
print filename, "started..."
curMachine = filename.split('/')[-1][:-4]
# curMachine = "cpu" + filename.split('/')[-1].strip('.csv')
model = Rnn_model.Rnn_model(data=data, machineID = curMachine, eta=curEta, lmda=curLmda)
model.fit()
pred = []
lastFc = None
for p in range(input_window, len(data)-predic_window,predic_window):
fc = model.predict(predic_window)
if lastFc is not None:
fc[0] = lastFc
lastFc = fc[-1]
fc[fc<0] = minimum
pred.append(fc)
model.update()
pred = np.array(pred).ravel()
f = filename.split('/')[-1]
fileutils.writeCSV("d:/data/cpu_rnn_forecasts/"+f, np.atleast_2d(pred))
print filename, "complete!"
def performsSlidingWindowForecast(filename,
minpercentile=5,
step=30,
input_window=3000,
predic_window=30):
'''
Input window = 250 hours = 250*12 = 3000
look ahead window 60 samples = 5 hours = 720min/5 = 60
'''
data = np.genfromtxt(filename)
data = data / np.max(data)
minimum = np.percentile(data, minpercentile)
N = len(data)
result = []
print filename, "started..."
y = data[0:input_window].tolist()
model = HW_model.HW_model(y, minimum, 'additive')
for strIndex in range(0, N - input_window - predic_window, step):
if strIndex == 0:
model.fit()
else:
y = data[strIndex:strIndex + input_window].tolist()
model.update(y)
y_pred = model.predict(fc=predic_window)
result.append(y_pred)
f = filename.split('/')[-1]
fileutils.writeCSV("D:/Wikipage data/pageviews_hw/" + f,
np.atleast_2d(result))
print filename, "complete!"
def extractMachineData():
datafiles = fileutils.getFilelist(
"D:/googleClusterData/clusterdata-2011-1/task_usage")
machineUsage = {}
startAt = 0
for machine in machines:
machineUsage[machine] = []
for datafile in datafiles[startAt:]:
print datafile
for row in fileutils.getCsvRows(datafile):
curMachine = row[4]
if curMachine in machines:
machineUsage[curMachine].append(row)
for machine in machineUsage.keys():
if startAt == 0:
fileutils.writeCSV("d:/data/perMachine5/" + machine + ".csv",
machineUsage[machine])
else:
fileutils.writeCSV("d:/data/perMachine5/" + machine + ".csv",
machineUsage[machine],
mode='ab')
startAt += 1
machineUsage.clear()
for machine in machines:
machineUsage[machine] = []
def performsSlidingWindowForecast(filename, minpercentile=5, step=30, input_window=3000, predic_window=30, order_=1):
'''
Input window = 250 hours = 250*12 = 3000
look ahead window 60 samples = 5 hours = 720min/5 = 60
'''
# data = np.genfromtxt(filename)
# data = data/np.max(data)
data = np.nan_to_num(np.genfromtxt(filename, delimiter=',', skip_header=1)[:,1]).ravel()
minimum = np.percentile(data,minpercentile)
N = len(data)
result = []
max = np.max(data)
print filename, "started..."
for strIndex in range(0,N-input_window - predic_window, step):
if strIndex == 0:
y = data[:input_window]
model = Markov_model.Markov_model(y, maximum=max, order=order_)
model.fit()
else:
# y = data[strIndex:strIndex+input_window,1]
y = data[input_window + strIndex - step:input_window + strIndex]
model.update(y)
y_pred = model.predict(predic_window)
y_pred[y_pred<0] = minimum
result.append(y_pred.ravel())
f = filename.split('/')[-1]
fileutils.writeCSV("d:/data/cpu_markov"+str(order_)+"/"+f, np.atleast_2d(result))
print filename, "complete!"
def readAndAggregate(filename, outputDir, resource='cpu'):
colomn = 1 # start time
if resource == 'cpu':
colomn = 5
elif resource == 'memory':
colomn = 7
elif resource == 'diskIO':
colomn = 11
resourcePerTask = np.genfromtxt(filename,
delimiter=',',
skiprows=0,
usecols=(0, colomn),
filling_values='0')
fileCsv = filename.split('/')[-1]
strTime = 600e6
endTime = strTime + 300e6
globalEndTime = 2506200000000
numberOfRows = globalEndTime / 300e6
aggregatedData = np.zeros([numberOfRows, 2], dtype=float)
x = 0
aggregatedData[x, 0] = strTime
for row in resourcePerTask[:]:
time = np.float_(row[0])
if (time >= strTime and time < endTime):
row_f = np.float_(row[1])
aggregatedData[x, 1] += row_f
else:
strTime = endTime
endTime += 300e6
x += 1
aggregatedData[x, 0] = strTime
while x < numberOfRows:
strTime = endTime
endTime += 300e6
aggregatedData[x, 0] = strTime
x += 1
fileutils.writeCSV(outputDir + '/' + resource + '_' + fileCsv,
aggregatedData,
header=('Time', resource.capitalize()))
def readAndAggregate(filename, outputDir, resource='cpu'):
colomn = 1 # start time
if resource == 'cpu':
colomn = 5;
elif resource == 'memory':
colomn = 7
elif resource == 'diskIO':
colomn = 11
resourcePerTask = np.genfromtxt(filename, delimiter=',', skiprows=0, usecols=(0,colomn), filling_values = '0')
fileCsv = filename.split('/')[-1]
strTime = 600e6
endTime = strTime + 300e6
globalEndTime = 2506200000000
numberOfRows = globalEndTime/300e6
aggregatedData = np.zeros([numberOfRows, 2], dtype=float)
x = 0
aggregatedData[x,0] = strTime
for row in resourcePerTask[:]:
time = np.float_(row[0])
if (time>=strTime and time<endTime):
row_f = np.float_(row[1])
aggregatedData[x,1] += row_f
else:
strTime = endTime
endTime += 300e6
x +=1
aggregatedData[x,0] = strTime
while x < numberOfRows:
strTime = endTime
endTime += 300e6
aggregatedData[x,0] = strTime
x +=1
fileutils.writeCSV(outputDir+'/' + resource + '_' +fileCsv, aggregatedData, header=('Time',resource.capitalize()))
def performsSlidingWindowForecast(filename_eta_lmda,
minpercentile=5,
step=30,
input_window=3000,
predic_window=30):
'''
Input window = 250 hours = 250*12 = 3000
look ahead window 60 samples = 5 hours = 720min/5 = 60
'''
filename, curEta, curLmda = filename_eta_lmda
data = np.genfromtxt(filename, delimiter=',', skip_header=1, usecols=(1))
minimum = np.percentile(data, minpercentile)
N = len(data)
print filename, "started..."
curMachine = filename.split('/')[-1][:-4]
# curMachine = "cpu" + filename.split('/')[-1].strip('.csv')
model = Rnn_model.Rnn_model(data=data,
machineID=curMachine,
eta=curEta,
lmda=curLmda)
model.fit()
pred = []
lastFc = None
for p in range(input_window, len(data) - predic_window, predic_window):
fc = model.predict(predic_window)
if lastFc is not None:
fc[0] = lastFc
lastFc = fc[-1]
fc[fc < 0] = minimum
pred.append(fc)
model.update()
pred = np.array(pred).ravel()
f = filename.split('/')[-1]
fileutils.writeCSV("d:/data/cpu_rnn_forecasts/" + f, np.atleast_2d(pred))
print filename, "complete!"
# truevals = np.divide(truevals, np.max(truevals))
threshold = np.percentile(truevals, overload_percentile)
cur_results.append(eval.calc_RMSE(truevals[train_window:], forecasts))
for val in eval.calc_upper_lower_acc(truevals[train_window:], forecasts):
cur_results.append(val)
for val in eval.calc_persample_accuracy(truevals[train_window:], forecasts, threshold):
cur_results.append(val)
for val in eval.calc_overload_states_acc(truevals[train_window:], forecasts, threshold):
cur_results.append(val)
return cur_results
if __name__ == '__main__':
files = []
root = "d:/data/"+TYPE+"/"
for _, _, fs in os.walk(root):
for f in fs:
if f.endswith(".csv"):
files.append(f)
pool = ThreadPool(4)
# performEvaluations(files[0])
results = pool.map(performEvaluations, files)
pool.close()
pool.join()
fileutils.writeCSV("d:/data/results/"+TYPE+"_"+METHOD+".csv", results)
print METHOD+" "+ TYPE + " complete"
def performEvaluations(filename, train_window = 3000, overload_dur = 5, overload_percentile = 70, steps=30):
cur_results = []
forecasts = np.genfromtxt("d:/data/cpu_norm_forecasts/"+ filename,delimiter=',',usecols=range(0,steps)).ravel()
truevals = np.genfromtxt("d:/data/cpuRate/"+filename, delimiter=',',skip_header=1)[train_window:train_window+len(forecasts),1]
threshold = np.percentile(truevals, overload_percentile)
cur_results.append(eval.calc_RMSE(truevals, forecasts))
for val in eval.calc_upper_lower_acc(truevals, forecasts):
cur_results.append(val)
for val in eval.calc_persample_accuracy(truevals, forecasts, threshold):
cur_results.append(val)
for val in eval.calc_overload_states_acc(truevals, forecasts, threshold):
cur_results.append(val)
return cur_results
if __name__ == '__main__':
files = []
for _, _, fs in os.walk("d:/data/cpuRate/"):
for f in fs:
if f.endswith(".csv"):
files.append(f)
pool = ThreadPool(4)
results = pool.map(performEvaluations, files)
pool.close()
pool.join()
fileutils.writeCSV("d:/data/results/normaldist.csv", results)
def performsSlidingWindowForecast(params, minpercentile=5, training_window=30, input_window=3000, predic_window=30):
'''
Input window = 250 hours = 250*12 = 3000
look ahead window 60 samples = 5 hours = 720min/5 = 60
'''
filename, METHOD, TYPE, OUTPUT = params[0:4]
#Wikidata
# data = np.genfromtxt(filename)
# data = data/np.max(data)
lastFc = None
if TYPE == 'pageviews' or TYPE == 'network':
data = np.nan_to_num(np.genfromtxt(filename.replace(".csv",""))).ravel()
data = data/np.max(data)
else:
data = np.nan_to_num(np.genfromtxt(filename, delimiter=',', skip_header=1)[:,1]).ravel()
minimum = np.percentile(data,minpercentile)
N = len(data)
result = []
print filename, "started..."
for strIndex in range(0,N-input_window - predic_window, predic_window):
if strIndex == 0:
y = data[:input_window]
if METHOD == 'ar':
model = AR_model.AR_model(y, order=training_window)
elif METHOD == 'ma':
model = MA_model.MA_model(y,order=training_window)
elif METHOD == 'hw':
model = HW_model.HW_model(y, minimum, 'additive')
elif METHOD == 'markov1':
model = Markov_model.Markov_model(y, order=1)
elif METHOD == 'markov2':
model = Markov_model.Markov_model(y, order=2)
elif METHOD == 'press':
model = Press_model.Press_model(y)
elif METHOD == 'agile':
model = Wavelet_model.Wavelet_model(y)
elif METHOD == 'fnn':
filename, METHOD, TYPE, OUTPUT, INPUT, curEta, curLmda = params[:7]
curMachine = filename.split('/')[-1].replace(".csv",".xml")
if TYPE.startswith("memory"):
curMachine = curMachine.replace("memory", "cpu")
model = Fnn_model.Fnn_model(data=data, machineID = curMachine, netPath="../data/"+TYPE.replace("memory", "cpu")+"_networks/"+curMachine, eta=curEta, lmda=curLmda)
else:
model = Fnn_model.Fnn_model(data=data, machineID = curMachine, netPath="../data/"+TYPE+"_networks/"+curMachine, eta=curEta, lmda=curLmda)
elif METHOD == 'rnn':
filename, METHOD, TYPE, OUTPUT, INPUT, curEta, curLmda = params[:7]
curMachine = filename.split('/')[-1]
if TYPE.startswith("memory"):
curMachine = curMachine.replace("memory", "cpu")
model = Rnn_model.Rnn_model(data=data, machineID = curMachine, netPath="../data/"+TYPE.replace("memory", "cpu")+"_networks/"+curMachine.replace(".csv",".xml"), eta=curEta, lmda=curLmda)
else:
model = Rnn_model.Rnn_model(data=data, machineID = curMachine, netPath="../data/"+TYPE+"_rnn_networks/"+curMachine.replace(".csv",".xml"), eta=curEta, lmda=curLmda)
elif METHOD == 'entwine':
filename, METHOD, TYPE, OUTPUT, INPUT, curEta, curLmda = params[:7]
curMachine = filename.split('/')[-1]
data2 = np.nan_to_num(np.genfromtxt(filename.replace("cpu", "memory"), delimiter=',', skip_header=1)[:,1]).ravel()
model = Entwine_model.Entwine_model([data, data2], machineID = curMachine, netPath="../data/entwine_networks/"+curMachine.replace(".csv",".xml"), eta=curEta, lmda=curLmda)
model.fit()
else:
if METHOD == 'press':
y = data[input_window + strIndex - predic_window:input_window + strIndex]
else:
y = data[strIndex:strIndex+input_window]
model.update(y)
p = model.predict(predic_window)
y_pred = np.atleast_2d(p)
y_pred = np.reshape(y_pred, (predic_window,1))
if METHOD == 'rnn':
if lastFc is not None:
y_pred[0,0] = lastFc
lastFc = y_pred[-1,0]
y_pred[y_pred[:,0]<0,0] = minimum
result.append(y_pred[:,0])
f = filename.split('/')[-1]
fileutils.writeCSV(OUTPUT+TYPE+"_"+METHOD+"/"+f, np.atleast_2d(result))
print filename, "complete!"
def main():
global METHOD
global TYPE
global OUTPUT
global INPUT
# Launch main menu
# main_menu()
print "Comparing Forecasting methods:\n"
print "Please complete the following:"
print "Enter the type of data used in the evaluation:"
TYPE = raw_input(" >> ")
print "The base INPUT directory is:", INPUT+TYPE
print "and OUTPUT directory is:", OUTPUT+TYPE
print "Would you like to change it? y or n"
ch = raw_input(" >> ").lower()
if ch == 'y':
print "Enter the base path for the INPUT directory (without the type):"
INPUT = raw_input(" >> ")
if not os.path.isdir(INPUT):
print "Error: Please try again; INPUT directory:"
INPUT = raw_input(" >> ")
print "And enter the base path for evaluation OUTPUT directory (without the type):"
OUTPUT = raw_input(" >> ")
if not os.path.isdir(OUTPUT):
print "Error: Please try again; evaluation OUTPUT directory:"
OUTPUT = raw_input(" >> ")
#########
print "Please choose a method to evaluate:"
print "1. Holt-Winters"
print "2. Auto-regression"
print "3. 1st Markov chain"
print "4. 2nd Markov chain"
print "5. PRESS"
print "6. Agile"
print "7. FFNN Model"
print "8. RNN Model"
print "9. Entwine Model"
print "10. Moving Average"
print "11. Average combo4 Model"
print "12. FFNN combo4 Model"
print "13. Weighted Average model"
print "0. Quit"
choice = raw_input(" >> ")
ch = choice.lower();
if ch == '':
menu_actions['main_menu']()
elif ch == '0':
exit()
else:
METHOD = methods_dict[ch]
pool = ThreadPool(4)
files = fileutils.getFilelist(INPUT+TYPE)
params = []
if METHOD =='fnn':
if TYPE.startswith("memory"):
hyperpath = "../data/"+TYPE.replace("memory", "cpu")+"_networks/hyperparams.csv"
else:
hyperpath = "../data/"+TYPE+"_networks/hyperparams.csv"
hyperparms = np.genfromtxt(hyperpath, delimiter=',', dtype=None)
for curRow in hyperparms:
if TYPE.startswith("memory"):
params.append([INPUT+TYPE+'/'+curRow[0].replace("cpu", "memory").strip("'")+".csv", METHOD, TYPE, OUTPUT, INPUT, curRow[3], curRow[4]])
else:
params.append([INPUT+TYPE+'/'+curRow[0].strip("'")+".csv", METHOD, TYPE, OUTPUT, INPUT, curRow[3], curRow[4]])
elif METHOD =='rnn':
if TYPE.startswith("memory"):
hyperpath = "../data/"+TYPE.replace("memory", "cpu")+"_rnn_networks/hyperparams.csv"
else:
hyperpath = "../data/"+TYPE+"_rnn_networks/hyperparams.csv"
hyperparms = np.genfromtxt(hyperpath, delimiter=',', dtype=None)
for curRow in hyperparms:
if TYPE.startswith("memory"):
params.append([INPUT+TYPE+'/'+curRow[0].replace("cpu", "memory").strip("'")+".csv", METHOD, TYPE, OUTPUT, INPUT, curRow[3], curRow[4]])
else:
params.append([INPUT+TYPE+'/'+curRow[0].strip("'")+".csv", METHOD, TYPE, OUTPUT, INPUT, curRow[3], curRow[4]])
elif METHOD == 'entwine':
hyperpath = "../data/entwine_networks/hyperparams.csv"
hyperparms = np.genfromtxt(hyperpath, delimiter=',', dtype=None)
for curRow in hyperparms:
params.append([INPUT+TYPE+'/'+curRow[0].strip("'")+".csv", METHOD, TYPE, OUTPUT, INPUT, curRow[3], curRow[4]])
else:
for f in files:
params.append([f, METHOD, TYPE, OUTPUT, INPUT])
if METHOD == 'avg4' or METHOD == 'combo4' or METHOD == 'wa':
# ensembleModel(params[0])
pool.map(ensembleModel,params)
pool.close()
pool.join()
else:
# print "skip"
# performsSlidingWindowForecast(params[0])
pool.map(performsSlidingWindowForecast, params)
pool.close()
pool.join()
pool = ThreadPool(4)
results = pool.map(performEvaluations, params)
pool.close()
pool.join()
fileutils.writeCSV(OUTPUT+"results/"+TYPE+"_"+METHOD+".csv", results)
print METHOD+" "+ TYPE + " complete"
exit()
def ensembleModel(params, types=['ma','ar','fnn','agile'], step=30, input_window=3000):
input_size = len(types)
filename, METHOD, TYPE, OUTPUT = params[0:4]
filename = filename.split('/')[-1]
filename, METHOD, TYPE, OUTPUT = params[0:4]
filename = filename.split('/')[-1]
combine_model = np.genfromtxt(OUTPUT+TYPE+"_"+types[0]+"/"+filename, delimiter=',', usecols=range(0,30)).ravel()
truevals = np.nan_to_num(np.genfromtxt(OUTPUT+TYPE+"/"+filename, delimiter=',',skip_header=1, usecols=(1))[:input_window+len(combine_model)])
for t in types[1:]:
forecasts = np.genfromtxt(OUTPUT+TYPE+"_"+t+"/"+filename, delimiter=',', usecols=range(0,30)).ravel()
combine_model = np.vstack((combine_model, forecasts))
average_fc = np.average(combine_model, axis=0)
if METHOD == 'avg4':
fileutils.writeCSV(OUTPUT+TYPE+"_"+METHOD+"/"+filename, np.atleast_2d(average_fc).reshape([178,30]))
print filename, "complete"
return
if METHOD == 'combo4' or METHOD =='wa':
training = SupervisedDataSet(input_size, 1)
for i in range(step):
training.appendLinked([combine_model[t][i] for t in range(input_size)], truevals[i+input_window])
besterr = eval.calc_RMSE(truevals[input_window:input_window+step], average_fc[:step])
bestNet = None
for i in range(50):
if METHOD == 'wa':
net = buildNetwork(input_size, 1, hiddenclass=LinearLayer, bias=False)
else:
net = buildNetwork(input_size, 2, 1, hiddenclass=LinearLayer, bias=False)
trainer = BackpropTrainer(net, training, learningrate=0.001, shuffle=False)
trainer.trainEpochs(100)
err = eval.calc_RMSE(truevals[input_window:input_window+step], net.activateOnDataset(training))
if err < besterr:
bestNet = net
break
combo_fc = average_fc[0:step].tolist()
# combo_fc = []
if bestNet == None:
combo_fc = average_fc
else:
for i in range(step, len(combine_model[0]), step):
training = SupervisedDataSet(input_size, 1)
for j in range(i,i+step):
combo_fc.append(bestNet.activate([combine_model[t][j] for t in range(input_size)])[0])
training.appendLinked([combine_model[t][j] for t in range(input_size)], truevals[j+input_window])
trainer = BackpropTrainer(bestNet, training, learningrate=0.01, shuffle=False)
trainer.trainEpochs(2)
result = np.atleast_2d(combo_fc).reshape([178,30])
minimum = np.percentile(truevals,5)
result[0,result[0,:] < minimum] = minimum
fileutils.writeCSV(OUTPUT+TYPE+"_"+METHOD+"/"+filename, result)
print filename, "complete"
def performEvaluations(filename, train_window = 3000, overload_dur = 5, overload_percentile = 70, steps=30):
cur_results = []
forecasts = np.genfromtxt("d:/data/diskio_ar_forecasts/"+ filename,delimiter=',',usecols=range(0,steps)).ravel()
truevals = np.genfromtxt("d:/data/diskio/"+filename, delimiter=',',skip_header=1)[train_window:train_window+len(forecasts),1]
threshold = np.percentile(truevals, overload_percentile)
cur_results.append(eval.calc_RMSE(truevals, forecasts))
for val in eval.calc_upper_lower_acc(truevals, forecasts):
cur_results.append(val)
for val in eval.calc_persample_accuracy(truevals, forecasts, threshold):
cur_results.append(val)
for val in eval.calc_overload_states_acc(truevals, forecasts, threshold):
cur_results.append(val)
return cur_results
if __name__ == '__main__':
files = []
for _, _, fs in os.walk("d:/data/diskio/"):
for f in fs:
if f.endswith(".csv"):
files.append(f)
pool = ThreadPool(4)
results = pool.map(performEvaluations, files)
pool.close()
pool.join()
fileutils.writeCSV("d:/data/results/diskio_autoregressive.csv", results)
usecols=range(0, 30))).ravel() #
# truevals = np.genfromtxt("d:/Wikipage data/"+TYPE+"/"+filename, delimiter=',',skip_header=1)[:train_window+len(forecasts),1]
truevals = np.genfromtxt("d:/Wikipage data/" + TYPE + "/" +
filename)[:train_window + len(forecasts)]
truevals = truevals / np.max(truevals)
cur_results.append(eval.calc_RMSE(truevals[train_window:], forecasts))
for val in eval.calc_upper_lower_acc(truevals[train_window:], forecasts):
cur_results.append(val)
return cur_results
if __name__ == '__main__':
files = []
pool = ThreadPool(4)
root = "d:/Wikipage data/" + TYPE + "/"
for _, _, fs in os.walk(root):
for f in fs:
files.append(f)
# performEvaluations(files[0])
results = pool.map(performEvaluations, files)
pool.close()
pool.join()
fileutils.writeCSV(
"d:/Wikipage data/results/" + TYPE + "_" + METHOD + ".csv", results)
print METHOD + " " + TYPE + " complete"
def performEvaluations(filename, train_window = 3000, overload_dur = 5, overload_percentile = 70, steps=30):
cur_results = []
forecasts = np.genfromtxt("d:/data/diskio_hw_forecasts/"+ filename,delimiter=',',usecols=range(0,steps)).ravel()
truevals = np.genfromtxt("d:/data/diskio/"+filename, delimiter=',',skip_header=1)[train_window:train_window+len(forecasts),1]
threshold = np.percentile(truevals, overload_percentile)
cur_results.append(eval.calc_RMSE(truevals, forecasts))
for val in eval.calc_upper_lower_acc(truevals, forecasts):
cur_results.append(val)
for val in eval.calc_persample_accuracy(truevals, forecasts, threshold):
cur_results.append(val)
for val in eval.calc_overload_states_acc(truevals, forecasts, threshold):
cur_results.append(val)
return cur_results
if __name__ == '__main__':
files = []
for _, _, fs in os.walk("d:/data/diskio/"):
for f in fs:
if f.endswith(".csv"):
files.append(f)
pool = ThreadPool(4)
results = pool.map(performEvaluations, files)
pool.close()
pool.join()
fileutils.writeCSV("d:/data/results/diskio_holtwinters.csv", results)
def performEvaluations(filename, train_window = 3000, overload_dur = 5, overload_percentile = 70, steps=30):
cur_results = []
forecasts = np.genfromtxt("d:/data/diskio_hw_forecasts/"+ filename,delimiter=',',usecols=range(0,steps)).ravel()
truevals = np.genfromtxt("d:/data/diskio/"+filename, delimiter=',',skip_header=1)[train_window:train_window+len(forecasts),1]
threshold = np.percentile(truevals, overload_percentile)
cur_results.append(eval.calc_RMSE(truevals, forecasts))
for val in eval.calc_upper_lower_acc(truevals, forecasts):
cur_results.append(val)
for val in eval.calc_persample_accuracy(truevals, forecasts, threshold):
cur_results.append(val)
for val in eval.calc_overload_states_acc(truevals, forecasts, threshold):
cur_results.append(val)
return cur_results
if __name__ == '__main__':
files = []
for _, _, fs in os.walk("d:/data/diskio/"):
for f in fs:
if f.endswith(".csv"):
files.append(f)
pool = ThreadPool(4)
results = pool.map(performEvaluations, files)
pool.close()
pool.join()
fileutils.writeCSV("d:/data/results/diskio_holtwinters.csv", results)
delimiter=',',
skip_header=1)[train_window:train_window +
len(forecasts), 1]
threshold = np.percentile(truevals, overload_percentile)
cur_results.append(eval.calc_RMSE(truevals, forecasts))
for val in eval.calc_upper_lower_acc(truevals, forecasts):
cur_results.append(val)
for val in eval.calc_persample_accuracy(truevals, forecasts, threshold):
cur_results.append(val)
for val in eval.calc_overload_states_acc(truevals, forecasts, threshold):
cur_results.append(val)
return cur_results
if __name__ == '__main__':
files = []
for _, _, fs in os.walk("d:/data/cpuRate/"):
for f in fs:
if f.endswith(".csv"):
files.append(f)
pool = ThreadPool(4)
results = pool.map(performEvaluations, files)
pool.close()
pool.join()
fileutils.writeCSV("d:/data/results/normaldist.csv", results)
